Motivated by the current requirements of digital 3D museums in the low bandwidth networks, we present a novel and efficient approximation algorithm based on axial symmetry of 3D pottery model. Available simplification algorithms suppress detailed features of the mesh without any change to the rest of 3D model. In this paper, we reduce data for mesh representation while preserving the geometric approximation as well as the model quality of the resulting mesh. First, the main symmetry axis of the pottery model is determined and then main body is detected by slicing parallel planes and separation criteria introduced in our proposed algorithm. Second, actual handles are detected based on sector slicing planes using a robust handle separation scheme. Third, every detected part, i.e. main body and handles, is approximated using a novel circle fitting method. Finally, generated vertices are remeshed and create approximated mesh. Experimental results are presented to illustrate superiority and affectivity of our method. Compared with available mesh simplification algorithms and using the same amount of data to represent the model, the proposed approach gives significant improvement in the accuracy of the approximated 3D potteries.